Mh. England et Ac. Hirst, CHLOROFLUOROCARBON UPTAKE IN A WORLD OCEAN MODEL .2. SENSITIVITY TO SURFACE THERMOHALINE FORCING AND SUBSURFACE MIXING PARAMETERIZATIONS, J GEO RES-O, 102(C7), 1997, pp. 15709-15731
Part 1 of this study [England et al., 1994] examined the sensitivity o
f simulated oceanic chlorofluorocarbon (CFC) to changes in the way the
air-sea gas exchange rate is parameterized in a World Ocean model. In
part 2 we consider more closely the role of surface thermohaline forc
ing and subsurface mixing parameterizations in redistributing CFC-11 a
nd CFC-12 in the ocean. In particular, a series of five different mode
l ocean experiments are forced with the same air-sea CFC flux paramete
rization. The five cases include (1) a control run with a standard sea
sonal cycle in surface forcing and traditional Cartesian mixing, (2) a
run in which the production rate and salinity of Antarctic Bottom Wat
er (AABW) is increased, (3) a run in which the production, outflow rat
es, and density of North Atlantic Deep Water (NADW) is increased, (4)
a run with enhanced isopycnal mixing of passive tracers, and finally (
5) a run in which the effects of eddies on the mean ocean flow are par
ameterized following Gent and McWilliams [1990]. The simulated CFC upt
ake in the Southern Ocean far exceeds observations in the first four e
xperiments. The excessive uptake is linked to the poor model simulatio
ns of Southern Ocean deep water masses, where, for example, model Circ
umpolar Deep Water is typically 0.2 to 0.4 kg m(-3) too buoyant. The i
nsufficient density of the deep water allows for extensive penetration
of convective adjustment to great depth during winter, in contrast to
observations, and this results in excessive downward mixing of the CF
C-enriched surface waters. Compared with the control experiment, the S
outhern Ocean CFC uptake is reduced in the cases with increased AABW s
alinity and NADW density, as a result of slightly higher deep water de
nsity and reduced wintertime convection in those experiments. Neverthe
less, CFC uptake in the Southern Ocean still substantially exceeds obs
erved ocean CFC content in the adjusted surface forcing cases. The mos
t extreme uptake occurs in case 4, where, in addition to deep convecti
ve mixing of CFC, there is also mixing into the ocean interior along i
sopycnal surfaces having an unrealistic orientation. The Southern Ocea
n CFC uptake in case 5, using the mixing scheme of Gent and McWilliams
[1990], is dramatically reduced over that in the other runs. Only in
this run do deep densities approach the observed values, and wintertim
e convection is largely suppressed south of the Antarctic Circumpolar
Current. Deep penetration of CFC-rich water occurs only in the western
Weddell and Ross Seas. This run yields CFC sections in the Southern O
cean which compare most favourably with observations, although substan
tial differences still exist between observed and simulated CFC. The s
imulation of NADW production is problematic in all runs, with the CFC
signature indicating primary source regions in the Labrador Sea and im
mediately to the southeast of Greenland, while the Norwegian-Greenland
Sea overflow water (which is dominant in reality) plays only a minor
role. Lower NADW is insufficiently dense in all runs. Only in the run
with surface forcing designed to enhance NADW production does the CFC
signal penetrate down the western Atlantic boundary in a realistic man
ner. However, this case exhibits an unrealistic net ocean surface heat
loss adjacent to Greenland and so cannot be advocated as a technique
to improve model NADW production. Conventional depth sections and volu
metric maps of CFC concentration indicate that on the decadal timescal
es resolved by CFC uptake the dominant determining factor in overall m
odel ventilation is the choice of subsurface mixing scheme. The surfac
e thermohaline forcing only determines more subtle aspects of the subs
urface CFC content. This means that the choice of subgrid-scale mixing
scheme plays a key role in determining ocean model ventilation over d
ecadal to centennial timescales. This has important implications for c
limate model studies.